TW525078B - Image processing device, method and providing media - Google Patents

Abstract

The present invention can use the simple structure and low cost device for image processing. The main CPU 31 uses the graphic commands to conduct the convolution filter processing on the commands of the graphic engine 41. The graphic engine 41, in response of this command, repetitively conduct the following process: multiplying the pixel data in the texture area of the image memory 43 with the value of convolution filter coefficients and rewriting into the graphic area; then, conducting the convolution filter process.

Description

525078 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs V. Invention Description (1) A7 B7

The present invention relates to an image processing device, method and providing media, and more particularly to an image for image processing with a simple structure and low cost. Processing device, method and providing medium. [Background Art] FIG. 1 shows a configuration example of an image processing apparatus. 4 The video camera 1 captures a picture object ′ which is not shown in the figure, and the graphics sections 2 and 5 of the picture object. The image processing unit 2 stores ^: 枓 to the image, and stores the image data input by the video camera 1 into the memory U. CPIJ5 pre-calculates the operation of the private R box South " operation unit 12 and executes 12 = the image processing unit 2 and 4 12 opposes this command, and performs predetermined operations on the image data stored in the memory U The data of,: is output to the graphics processing unit 3. The graphics processing unit 3 uses the computational data input by the image processing unit 2 to save the memory of the graphics processing unit 3 by the CPU 2 and CPU 5: 22, and generates the display data from the computing data stored in the memory 21. = The display data generated by the graphics processing unit 3 is output to the CRT 4 for display. The CPU 5 also accepts the data output from the video camera 1 or the image processing unit 2 and performs predetermined processing as necessary to output to the image processing unit I; the graphics processing unit 3. In this image processing device, such as a camera! The output image data "is processed twice. For the second image by Beko, she is processed by convolution, and output to c R τ 4 to make her mind 'the action at that time is basically processed by the CPU 5 as shown in Figure 2. This CPU 5 is in step S1, and the tk video camera 1 accepts a picture of 1 frame. -------------- Install --- (Please read the back first, note the matter and write (This page) Order:; Line ·

525078 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economy V. Invention Description (2) Material supply. This image data is used as the image data of the root frame. For example, as shown in FIG. 3 ′, it is composed of ΗMAX × VMAX pixel data Csp. At step 骡 S1, the CPU 5 also sets the convolution filter coefficient c [m] [η]. This convolution filter coefficient .Cv [m] [η] is composed of 3 × 3 coefficients in the case of the example in FIG. 4. Next, in step S2, the CPU 5 regards the pixel data cdp of the target frame as shown in FIG. 5 and assumes ΜΜAX × VMAX pixel data, and the pixel data Cdp [l] [of the coordinates (1, 1) therein i] The value is initially set to zero. At step S3, 1 is initially set at variable j, and at step S4, 1 is set at variable i. This variable i, j is shown in Fig. 3 and Fig. 5 and represents the horizontal coordinate ⑴ or the vertical coordinate (j,) of the image of the source frame (Figure 3) and the image of the target frame (Figure 5). i takes a value from 0 to HMAX-1, and j takes a value from 0 to vmAX-1. In step S5 and step S6, 0 is initially set at the variable η and the variable m, respectively. The variables m and η are shown in Fig. 4 and represent the horizontal coordinate (m) and the vertical coordinate (η) of the convolution filter coefficient. In this case, m and η both take values from 0 to 2. Next, in step S7, the CPU 5 performs the operation shown in the following formula: Cdp [i] [j] = Cdp [i] [j] + Cv [m] [n] * CSP [i + m-1] [j + n-1] (1) Now let j = l, i = l, n = 0, m = 0, and let Cdp [i] [j] start value be 0, so we can get the following formula: Cdp [1 ] [1] = Cv [〇] [0] * CSp [〇] [0] (2) _____- 5- This paper size applies to China National Standard (CNS) Α ^^ Γ (210, X297 mm) I " '-(Please read the ins and outs on the back of this page first) • Binding and Binding Line $ 25078 A7 R7 --------- ------ 5. Description of the Invention (3) Next Go to step S8, the CPU 5 determines whether the variable m is smaller than two. In the present case, m = 0, which is smaller than 2, so it proceeds to step S9, m only increases by ^, and is set to hi = 1. Then, it returns to step S7 and performs the calculation of the above-mentioned expression (丨). As a result, the following formula can be obtained:

Cdp [1] [1] = Cdp [1] [1] + Cv [1] [0] * CSP [1] [0] (3) Also 'here Cdp [1] [ 1] 値 is 値 obtained by the above formula (2). Then, it proceeds to step S8, and determines whether the variable πm 疋 is smaller than 2. In the present case, m, = 1, so it proceeds to step s9, and the variable m is increased by only 1 and said to be m = 2. After that, return to step S7, and then calculate the above (1) to obtain the following formula:

Cdp [1] [1]-Cdp [1] [1] + Cv [2] [0] * CSP [2] [0] (4) Process the 'root source pixel data Csp [〇] [〇] according to the above, Csp [1] [〇], csp [2] [〇] are multiplied by the convolution filter coefficients Cv [〇h〇], cv [ι] [0], cv [2] [0], and the cumulative multiplication can be obtained.値 値 値. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Next, proceed to step s8 to determine whether the variable m is smaller than two. In the current situation, m = 2, so whether to make a judgment is made, and the process proceeds to step S10. In step S10, it is determined whether η is smaller than 2. In the present case, n = 0, so a "yes" determination is made and the process proceeds to step S 1 1. In step s11, the variable η is only increased by 1, and in the present case, η = 1 is set. After 'return to step S6' after initial setting is m = 0 ', in step S7, 525078 A7 V. Description of the invention (4 Perform the operation of the above formula. Based on this, perform the following operation: Cdp [!] [!] = Cdp [1] [1] + Cv [0] [1]% Csp [〇] [i] Next, go to step S8 to determine whether the variable m is more than two. 〖: In the current situation, m = G, so the process is To judge, cut S9, set it to m: and then return to step S7, and then perform the following formula Cdp [1] [1] = Cdp [!] [!] + (： Ν [ 1] [1;] ^ Csp Π] [1] (6) Repeat the results of the above process, and perform the following formulas to (i〇): Cdp [1] [1] = Cdp [l] [i] + Cv [2] [1] * cSP [2] [1] (Please read the phonetic on the back? Write this page first)

Printed by Tfe'JA, Employee Consumer Cooperative of Intellectual Property Bureau, Ministry of Economic Affairs ⑺

Cdp [1] [1] = Cdp [1] [1] -fCv [0] [2] * Csp [〇] [2] (8)

Cdp [1] [1] = Cdp [1] [1] + Cv [1] [2] * CSP [i] [2] (9)

Cdp [1] [1] = Cdp [1] [1] + Cv [2] [2] * CSP [2] [2] (10) Based on this, 1 pixel data CSP [1] [1] The convolution filtering for the object pixels is processed. At this time, 'at step S10', 11 = 2, so it is judged whether to go to step S12. In step S12, it is determined whether the variable i is smaller than HMAX-2 (in the case of this example, HMAX-2 is 6, so HMAX-2 = 4). The current situation, i = 1 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 525078 Ministry of Intelligent Finance A7 V. Invention Description (5) 'Small than HMAX-2, so If yes, go to step si 3, add 1 /, and add 1 to set 1 = 2. After that, it returns to step S5 and executes the subsequent steps! . That is, the convolution filtering operation using the pixel data CSP⑴⑴ as the target pixel is performed in the same manner as the convolution filtering operation with the pixel data CSP⑴⑴ as the target pixel. Convolution filtering processing operations on pixel data according to the following list: 'Enter the pixel data Csp [1] [HMAX_2] (in the case of the example in FIG. 3, ~ [1] [4])' In step S12, a determination is made as to whether or not to proceed, and the process proceeds to step S12. In step su, it is determined whether the variable j is smaller than νΜΑχ_2 (in the case of this example, VMAX = 6, so νΜΑχ_2 = 4). The current situation is M, so the judgment is yes, and the process proceeds to step S15, where "i" is incremented only and "2" is set. After that, it returns to step S4 and executes subsequent processing. That is, the convolution filtering processing operation on the pixel data in the j = 2 range is performed in the same manner as described above. As above, the convolutional transport office where j is a row of 4ΧΜΑχ_2 in the column of νΜAX × 2 for = picture element data Csp_AX_2] [VMax_2] will perform data processing on the result obtained on the part Γ 円 5 for graphic processing. two. The graphics processing unit 3 uses the "quoted" image input from which 5 to the C memory 21. The display data generating section 22 converts this image data into non-depleted material and outputs it to the CRT 4 for display. In the above, the convolution filtering arithmetic processing is executed by the image processing unit 2. Or sometimes other =, but sometimes use dedicated hardware to make it execute. Feng Beima executed this type of printing ruler (21 () x 2 525.078 A7 B7 V. Description of invention (6 Issues to be printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs, which are to be solved by the poor) In the picture as above The image processing device uses the hardware 5 and the image processing unit ^ to perform predetermined calculations in hardware. As a result, the burden on processing 5 or 2 in Figure 2 becomes larger or special hardware is required. Not only the structure becomes complex but also costs. Increasing subject. The present invention has been completed in view of this situation, and has a simple structure and can perform image processing. [Means to Solve the Problem] • Apply for the image processing device included in item i of the f benefit scope, It is characterized by having a first and a fifth memory mechanism: memorizing the root image data in pixel units; a second memory mechanism: memorizing target image data in pixel units ... a drawing mechanism: repeatedly performing the following operations until a predetermined calculation result is obtained ·· For the second memory: the memory, the source of the structure of the image data is subject to predetermined operations in pixel units and riding in the polygon unit: the memory mechanism as the target image data. Contained image The second method is an image processing device that stores the image data of the source image in pixel units. , Which is characterized in that it includes the step of drawing a margin: repeatedly ordering the following actions to obtain a predetermined operation result: the memory is stored in the _th record, the source image data is subjected to a predetermined operation in pixel units, and the polygon unit is %% to the second The memory mechanism serves as the target image data. The provided medium described in item 9 of the scope of patent application is characterized by providing a program for causing the image processing device to execute processing including a drawing step, and the image processing device has a first-memory Organization ·· Remember the root image in pixel units -------------- install ------ (please read the note on the back to write this page) _ Order:-丨 线· In-525078 Description of the invention (Image data printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs; and, 筮-七 卜 立 ww Hanging out of the institution ... · Record the image data in pixel units, the depiction Step 希 $ 希, 彳, / k 目 & 0 q Steps are repeated The operation described above is to get the scheduled operation result.... "Recognized in the first modern product her technical recognition," 疋 连 # π scheduled «and" == special source image data in pixel units for the image data. ^ The early position rides on the second memory mechanism as the image processing device included in the item 1m of the crime. It is characterized by: It has a memory mechanism, an industry; ψ4, a health institution, and an enforcement agency. A memory department: using a map to set up the homepage, often thinking about the source image data; and, the second memory #taking a picture element ^^ to recall the target image data, and the generating mechanism generates a drawing ^ gas The 叩 π 叩 order is to make the following actions repeatedly until a predetermined calculation result is obtained + ^ Jiejie The first root image data of Yihan is pre-transported in pixel units and traced to the second memory in polygon units As the target image data, the execution unit executes the drawing order generated by the generation unit. The image processing method contained in the patent application No. 丨 7 is characterized by including a memory step: the root image data is memorized in the unit of pixel _ a in pixel units, and the target image is lean pixels. The unit is memorized in the second memory unit: and, generating step: generating a drawing command, which repeatedly performs the following actions to obtain a predetermined operation result: in the memory step, the root image data stored in the first memory unit is plotted The prime unit performs a predetermined operation and uses the polygonal trajectory to describe the target image data in the brother's own memory. The provided medium contained in item 18 of the scope of the patent application is characterized in that a program for executing a process including a memory step and a generation step is provided, and the memory step is to cause the root image data to be memorized in the first memory unit in pixel units. At the same time, the target image data is memorized in the second memory unit in pixel units. The generated paper size is in accordance with the Chinese National Standard (CNS) A4 specification (210 x 297 mm). C Please read the intentions on the back first Νξ W this page)-equipment; line five, description of the invention (8 to = generate a riding command, the drawing command system, the following actions are repeated to perform the root operation result: in the memory step to the memory stored in the first- For example, the pixel unit is used to calculate the data and the polygon unit is used to describe, "Yuan Wan; the second memory unit is the target image data. The image processing device contained in the patent _19 item is characterized by: Brother-memory mechanism: The prime unit memorizes the source image data; the first memory mechanism: the target unit data of the prime unit field; the first-drawing edge machine. The root-memory map stored in the first memory mechanism uses a part of the predetermined operation Transport ; 丨 ,, Xi, Zikou Leping 1 commit ^ ^ Μ ΑΛ:-^ Xi 圪 unit is depicted in the second record structure as the target image data structure: the root image data memorized in the brother-memory mechanism is illustrated The prime unit is given a reward: other—part of the calculation, plus or minus the image data that has been used by the first rider and drawn in polygonal units on the m—image data. The target picture applies for patent scope No. 26 The contained image processing method is characterized in that it includes a first drawing step: the recognition of poverty, the use of a root image in the memory of the younger one, and the operation of a subdivision in the juice. The polygon drawing at the second memory mechanism is used as the drawing step: the root image data stored in the second memory structure of the first-memory machine is applied to the other part of the predetermined operation in pixel units. The image data that has been traced in the first drawing step is added or subtracted, and the second memory mechanism is used as the target image data in polygon units. The provided media described in item 27 of the scope of the patent application is characterized by: Enhancement μ includes the first drawing and the second interpolation The execution procedure of the green step-11. J ^ 〇78 V. Description of the invention (9) The Department of Science is to memorize the memory in the first place—the memory mechanism applies the predetermined position of the manure φ Μ winter shellfish to the picture in the early position. -Memory mechanism = calculate and use polygon in a single step to memorize in the first—memory mechanism is operated by other parts of the predetermined operation. Project ^ Image data. The image processing described in Fanyuan " in Item 1 of the scope of patent application; == device, patent application, ... repeatedly perform the following operations until the predetermined calculation result is obtained : The image data of the root of Yudi's first memory organization is pre-funded in pixel units, and the second memory organization is described in polygonal units as the target image = the image contained in item 10 of the scope of patent application Like image processing device, patent application = # 7: The image processing method and the scope of patent application No. I8 are printed on clothing by the Consumers ’Cooperative of the Intellectual Property Bureau of Qibei. ^ Provide media to generate riding commands, the drawing commands are The following report: After the execution, the predetermined operation result is obtained: the image data stored in the first memory unit is calculated in pixel units and is exposed in polygon units, and the memory unit is used as the target image data. First, the image processing device described in item 19 of the scope of patent application, the image processing method described in item 26 of the patent application, and item 1 of the scope of patent application "1. Provide the media and perform the following actions: The root memory image of the first memory machine is given a part of the predetermined operation in pixel units, and is depicted in the second memory mechanism as the target image in rim units.

Ben ’': Known Scale Turning (CNS) A4 Specification (21G xlgf ^ TT 525078 A7 B7

V. Description of the invention (11) 'Working student drawing seal order, the drawing order is to repeatedly perform the following actions to obtain the pre-calculation result: the root image memorized in the first-memory part Xin == operation and The polygon unit description is read in the second memory as the target image; and, the execution mechanism (such as the drawing engine of FIG. 6: the execution order of the drawing command generated by the generation mechanism. 1 = the picture in the profit range item 19) The image processing device is characterized by having a brother-memory mechanism (for example, the texture area 51 in FIG. 7): memorizing the root image data; and a memory mechanism (for example, the riding area in FIG. 7 ^ two pixels, unit memory target image Image data; first tracing mechanism: step, step S38, step S39): The root image data stored in the first memory machine f is subjected to predetermined operations in pixel units— 单位 运. And Γ I : Early position is drawn in the second? Jiyi mechanism as the target image data 'and brother-drawing mechanism (for example, step s37 in Fig. 12 screams. · The source image data stored in the -memory mechanism is illustrated The culprit adds the image data of the other μ ^ in the pre-calculation operation. π ^ ^ ㈣ The tracing mechanism recalls the mechanism as the target image data holder. 乂 " ^ The edge is depicted in the image processing device contained in item 21 of the scope of the second patent application, which is characterized by ΓΓΓΓ (For example, step S34 in FIG. 12): a person who specifies a calculation mode between the source image frame and the target image data. [: A block diagram of a computer entertainment device H for displaying the image processing device to which the present invention is applied. Through the bus 34, the main memory 32 and the image processing chip 33 are connected to the main CPU 31. The main CPU 3I generates a riding command,

Order line (------- --__ 14 525078 A7 Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs Employee Consumer Cooperatives V. Invention description (12) Actions of the image processing chip 33. Appropriate memory κρυ3ι in the main memory 32 Programs and data required for executing various processes, etc. Image processing chip 33 < Insertion engine 41 corresponds to the 4U supplied by the main CPU 31 'is executed in the image memory 43 through the memory interface 42 An operation for drawing predetermined image data. The bus 45 is connected between the memory interface 42 and the rendering engine 41, and the bus 46 is connected between the memory interface 42 and the image memory 43. The bus 46 has 128, for example. The bit width of the bit 'drawing engine 41' can perform drawing processing at a high speed with respect to the image memory 43. The drawing engine 41 has, for example, the following capabilities: (reaitime) at (1/30 to 1/60, second ) Describing image data of 320 X 240 pixels, such as NTSC method, pAL method, or 64 × 480 pixels, between dozens to dozens of times. Image processing chip 33 is also programmable. Cathode Ray Tube Controller (PCRTC) 44 which has The following functions: real-time control of the position, size, resolution, etc. of the image data input by the video camera 35. The PCRTC 44 writes the image data input by the video camera 35 into the image record through the memory interface 42. The texture area 5 1 (Figure 7). In addition, the PCRTC 44 reads the image data drawn in the drawing area 52 (Figure 7) of the image memory 43 through the memory interface 46, and outputs this to the crt 36 to make it Display. As shown in FIG. 7, image 1 and memory 43 form a unified memory structure that can specify texture area 51 and tracing area 5 2 in the same area. The main CPU 3 1 5 When the predetermined texture is drawn in the drawing area 52, the "Generate FlatTexturePolygon" command, enter this -15- (Please read the precautions on the back first-install this page) Order-. • f · The paper size is suitable丨 ΗChinese National Standard (CNS) A4 specification (210 X 297 mm) 525078 A7 __B7 V. Description of the invention (13) Out to the drawing engine 41 1. For example, the command to draw a triangular polygon is the following command:

Flat Texture Triangle (DxO, DyO, Dxl, Dyi, j) 2

Dy2, SxO, SyO, Sxl, Syl. Sx2, Sy2, L) Here, DxO, DyO, Dxl, Dyl, Dx2, Dv, soil, • X uxz Uy2 represents the coordinates of the vertices of the triangle drawn on the target (drawing area 52) , SxO, Sy0, Sxl, Syl, Sx2, Sy2 represent the coordinates of the triangle apex of the root (texture area 51). In addition, L · represents a brightness 値 multiplied by a texture pixel 値 in a polygon (triangle) represented by a dot line (Sxn, Syn). Similarly the 'quadrilateral drawing command is as follows:

Flat Texture Rectangle (DxO, DyO, Dxl, Dyl, Dx2

Dy2, Dx3, Dy3, SxO, SyO, Sxl, Syl, Sx2, Sy2, Sx8, Sy8, Sy3, L) This is also DxO, DyO, Dxl, Dyl, Dx2, Dy2, Dyy 3 Will hit the quadrilateral vertex of the target (drawing area 5 2),

SxO, SyO, Sxl, Syl, Sx2, Sy2, Sx3, Sy3 represent the origin (texture area 5 1) quadrilateral vertex coordinates. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs -------------- Dish --- C Please read the notes on the back first and write this page) Draw the engine 4 1 from the main CPU 3 1 through the confluence For example, in row 34, if a triangle drawing command is input, the points (Sxn, Syn) in the polygon specified by the three vertex coordinates (Dx0, DyO), (Dxl, Dyl), and (Dx2, Dy2) in the texture area 51 are input. The result of multiplying the pixel 値 by 値 L is plotted at the corresponding points (Dxn, Dyn) within the polygon specified by the coordinates (Dx2, Dy2), (SxO, SyO), (Sxl, Syl). The coordinates (Dxn, Dyn) on the drawing area 52 corresponding to the coordinates (Sxn, Syn) of the texture area 51 can be expressed by the following formula (1 1). -16 The paper size is suitable for the Chinese National Standard (CNS) A4 size (2.10 X 297 mm) 525078

V. Description of the invention (14 affine transformation) [Number 1]

(11) In the above formula, a to d are coefficients for rotation, and 0FX and 0F. Are coefficients for parallel movement. In the drawing area 52, drawing is performed in units of pixels. For Tian Suzhen, the coordinate map element 的 on the texture area 51, which is obtained from the coordinates by using the inverse-fitting transformation expressed by the following formula (12), is used. [Number 2] / x + h 卜 厂 v Ί I Λ ~ ^ Λ, •• (12)

(Please read the caution page on the back first) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs ^^ In the above formula, h and v are coefficients of 0 or more and 1 · 0, and x, y, X, and Y are integers . In addition, this coordinate 値 usually has a decimal part. As shown in FIG. 8, it becomes a coordinate located between pixels. Therefore, bilinear interpolation is used for the pixel 値 SP (X , Y). SP (X, Y) = (1 — h) 氺 (1-v) 氺 TP (X, y) " + h 氺 (1 -V) 氺 TP (x + 1, y) + (1 — h)氺 v 氺 TP (x, y + 1) + h 氺 v 氺 TP (x + 1, y ten 1) (13) ______- 17 -_____ This paper rule is suitable for the Chinese National Standard (CNS) A4 specifications (210 x 297 mm) 525078 A7 B7 V. Description of the invention (15) That is, 'From the point TP (x, y) on the texture area 51 only leaves + h in the X-axis direction and leaves only the tangent point in the y-axis direction. The pixel 値 of (x + h, y + v) and the points Tp (x 'y), tp (x + i, y), tp (X, y 0, TP, y + 1) from its surroundings For distance, use the weighted 値. Further, the rendering engine 41 corresponds to a modal which is specified by the mixed modal setting function Set Mode (MODE) from the main CPU 31, and the target pixel 値 DP (X, Y) and texture on the rendering area 52 The root pixel 値 SP (X, Y) on the area 51 is subjected to a blending process. Among the mixed modes executed by the drawing engine 41, there are modes 0 to 3, and the following modes are performed in each mode: MODEO · SP (x, γ) MODE1: DP (X, γ) + Sp (X, γ) MODE2: DP (X, γ) _ SF (X, γ) MODE3 · (1-asp (χ, γ)) * Dp (χ, γ) + aSP (X, Υ) * SF (X, γ) Also 'aSP (X' Υ) represents & 値 of the root pixel 値. That is, in the mode 0, the root pixel 値 is still drawn on the target, in the mode 1 the root pixel 値 is added to the target pixel 描绘, and in the mode 2 the root pixel 値 is subtracted from the target pixel 値Draw. Further, in mode 3, the source pixel 値 and the target pixel 値 are synthesized by weighting corresponding to the source a 値. The image data drawn in the drawing area 52 of the pixel memory 43 is transmitted to the pCRTC 44 by recording L, and the 46 items are output to the pCRTC 44, where they are output to% display. Next, after performing convolution filtering processing on the image data input by the video camera 35, the second processing is performed when the output is 36. ί Please read the note on the back π this page} Clothing printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

525078 A7 _B7_ V. Description of the invention (17) for (m = 0; m <3; m ++) {Flat-Texture-Rectangle (1, 1, HMAX-1, 1, HMAX-1, VMAX-1, 1, VMAX-1, m, n.HMAX-2 + m, n, HMAX-2 + m, VMAX-2 + n, m, VMAX-2 + n Cv [m] [n]): (Please read the first Attention printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

The above process is represented as a flowchart, as shown in FIG. 12. That is, first, in step S31, the main CPU 31 clears the drawing engine 52 as the target drawing area 52 (output area), and sets the convolution filter coefficient Cv [m] [η] to a predetermined value. In the present case, as shown in FIG. 11, 3 x 3 convolution filter coefficients are prepared. Next, in step S32, the main CPU 31 outputs a command to the rendering engine 41 to set the mode 0 as the mixed mode. In response to this command, the drawing engine 41 sets the mixed mode to the mode 0. Next, in step S33, the main CPU 31 generates a quadrangle (1, 1, HMAX-1, 1, HMAX-1, VMAX-1, 1, VMAX-1) among the targets composed of UMAX X VMAX pixels. The command for rendering pixels 値 0 is output to the rendering engine 41. The drawing engine 41 corresponds to this command. As shown in FIG. 13, among the targets of the UMAX X VMAX, the sketches of each drawing except for the row of j = 0 and the row of i = 0 are 0. Next, in step S34, the rendering engine 41 corresponds to a command from the main CPU 31, and sets the mode 1 as the hybrid mode. That is, the following modalities are fixed. -20- This paper size applies the Chinese National Standard (CNS) A4 specification (210x297 mm) Γ page) • Packing ·-^ > 25078 A7

5. Description of the invention (20), dpl2 ′ 〇〇 + CSpi〇 X Ci〇 + CSP20 x c2〇 + c SP01 It is judged whether to go to step S40. In step s40, it is determined whether η 疋 is smaller than 2. In the present case, n = 0, so proceed to YES. The variable n is only incremented by i, and is set to ㈣, and then go to step S36. In step S36, the variable calendar is reset to 0, and in step S37, the following processing is performed: multiplying the source map Cdp of the fan garden specified by Bu 0 to 4 and 1 to 4 by the convolution filter coefficient cv [〇] n] (= c0 丨), and then draw on the target that has been drawn. By this, for example, as shown in FIG. 17, Cdpn, Cdp, Cdp2i, and C Cdp22 become as follows: Cdpl 1 = CSP〇〇XC〇〇 + X Cn, c dp2i = csp 丨 0 X c00 + cSP20 X c10 + c X cc, 0 1 sp3〇x C20 + c SPl l dpi2 = CSP01 X c00 + Cspm X C10 + cSP21 X c20 + c ^! SP02

Cdp22 = Csph X C〇〇 + CSP21 X Cl0 + CSP31 x 〇

20 + C SP12 Employee Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs', the same processing will be repeatedly executed to the judgment in step S40: 2 _ ,. As a result, the target area is traced as shown in Figure i8:

read. As shown in the figure, for example, Cdpi becomes as follows. C dp 1 1-C s POO X C 〇〇 + C s P 1 0 X C 1 〇 + Q

dp, C

dP2l 、 C dp 1 c dp22

XC〇i + Cspn X Cii + CsP21 X C2 c SP20 X C20 + c SP02 X c02 + c spo SPl-system-23 corpses, adaptable 6 Chinese National Standard (CNS) M specifications (210 > < 297 mm) CC 07 5 2 5 A7 _B7_ V. Description of the invention (21) X C12 X + CsP22 X CsP22

Cdp2i = CSP10 x C00 CSP20 x C10 + CSP3〇 x C20 + cSP11 X C〇i + CsP21 X Ch + CsP31 X C21 + CSP12 X C〇2 + CsP22 X C] 2 X + c s p 3 2 X C s p 2 2

Cdpl2 = Csp〇l X C〇〇 + CSP11 X C10 + CSP21 X C20 + Csp〇2 X C〇i + CSP12 X Cn + CSP22 X C21 + Csp〇3 X C〇2 + CSp13 X C12 X + CsP23 X CsP22

Cdp22 = Cspil X Ch > 〇 + CSP21 X C10 + CSP31 X C20 + CSP12 X C01 + CSP22 X Cn + CsP32 X C2 丨 + CsP13 XC〇2 + CSP23 XC 1 2 X + csp 3 3 XC s P 2 2 In this case, as the root pixel data, cSP16 pixel 値 is needed, but this image data does not actually exist, so as shown in Figure 16, j = 5 (= VMAX-1) pixel data Cdp is Think of invalid pixel data. As described above, the pixel data for performing the convolution filtering process can be obtained in the ranges specified by i = 1 to HMAX-2 and j = 1 to VMAX-2. The above processing is the processing of the convolution filter coefficient timing, but when the convolution filter coefficient is negative, the rendering command output from the main CPU 31 to the rendering engine 41 is as follows:

Set_Mode (0);

Flat—Rectangle (1,1, HMAX-1,1, HMAX-1, VMAX-1,1, VMAX-1, CMAX / 2); for (n = 0; n <3; n ++) {for (m = 0; m <3; m ++) {if (Cv [m] [n] > 0 {-24- This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) (Please read the back first Note for the is-installation --- I write this page) Order: Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 525078 A7 B7 V. Description of Invention (22)

SetJViode (l);

Flat_Texture_Rectangle (1,1, HMAX-1,1, HMAX_1, VMAX-1,1, VMAX-1, m, n, HMAX-2 + m, η, HMAX-2 + m, VMAX-2 + n, m, VMAX-2 + n Cv [m] [n]):} if (Cv [m] [n] < 0)}

Set_Mode (2);

Flat_Texture_Rectangle (1,1, HMAX-1,1, HMAX-1, VMAX-1,1, VMAX-1, m, n, HMAX-2 + m, li, HMAX_2 + m, VMAX-2 + n, m, VMAX-2 + n -Cv [m] [n]):}

Set_Mode (2);

Flat—Rectangle (1, 1, HMAX-1, 1, HMAX-1, VMAX-1, 1, VMAX-1, CMAX / 2); In addition, this process is represented as a flowchart, as shown in Figure 19 and Figure 20 shown. That is, first, in step S51, the convolution filter coefficient Cv [m] [n] is set to a predetermined value. Next, in step S52, the rendering engine 41 corresponds to a command from the main CPU 31, and the mode 0 is also set as the mixed mode. Then, in step S53, the rendering engine 41 renders CMAX on the target area quadrilateral (1, 1, HMAX-1, 1, HMAX-1, VMAX-1, 1, VMAX-1) in the drawing area 52 of the image memory 43. /2. CMAX means the maximum value of the pixel 値. That is, in this case, as shown in FIG. 21, the pixel 値 maximum 値 1/2 値 is drawn in a range defined by i = 1 to HMAX-1 and VMAX-1 in the target area. -25- The size of this paper is suitable for the national standard of the country (CNS) A4 (210 X 297 mm)

(Please read the note at the back of this page K W) Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

V. Description of the invention (23) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. Next, at step S54 and step S55, the variables η and m are initially set to 0. In step S56, the convolution filter coefficient Cv [mnn is determined. ] Is it positive. When the filter coefficient is positive, proceed to step s 5 7 and set the mode 1 as the mixed mode. Next, it progresses to step S58 and performs drawing processing. This drawing process is the same process as in the case of step S 3 7 in FIG. 12 described above, so its description is omitted. The processing of step S 58 is followed by step S 59. It is judged whether the variable m is smaller than 2 and it is judged that m is longer than 2 hours. Then it proceeds to step S 60 and the variable m is only increased by 1 and the process returns to step S56. Then, the subsequent processing is executed repeatedly. In step S59, it is judged that the variable m is not longer than 2 hours, that is, when m and 2 are equal to each other, it proceeds to step S61 to determine whether the variable η is smaller than 2. The variable η is longer than 2 hours 'and the process proceeds to step S62. After the variable η is incremented by 1', the process returns to step S55, and the process thereafter is repeated. In step S61 1, it is determined that the variable η is not longer than 2 hours (the variables η and 2 are equal), and the process proceeds to step S63, and as the mixed mode, the mode 2 is set. Then, in step S64, the following processing is performed: the rendering engine 41 is subtracted from the pixel 値 of the target quadrangle (丨, 1'HMAX-1'1'HMAX-1, VMAX-1,1, vmax_i), which is depicted in step S53. CMAX / 2 has been described. On the other hand, in step S56, when it is judged that the convolution filter coefficient cv [m] [n] is not positive, it proceeds to step S65 to judge whether the coefficient is negative. When it is judged that the coefficient is negative, the process proceeds to step S66, and as the mixed mode, mode 2 is set. Then, in step S67, the same description processing as in the case of step S58 (ie, step S37 in FIG. 12) is performed. In step S65, when the judgment coefficient Cv [m] [n] is not negative, it is judged that this is ____- 26-This paper size applies the Chinese National Standard (CNS) A4 specification (21〇X 297 mm) '- -------- Order --------- (Please read the note on the back of this page first) 525078 Α7 Β7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs V. Invention Description (24) Number If it is 0, the processes of steps S66 and S67 are skipped. Then, the processing of step S67 is followed by the processing of step S59, and the same processing as that described above is executed. Concerning the above, it can be said that by drawing the green pixel data in a polygon unit, a convolution filtering operation process is performed, but the same processing may be performed in a pixel unit drawing. Fig. 22 shows the processing of this case. That is, first, in step S71, the convolution filter coefficient Cv [m] [n] is set. In step S72, as the mixed mode, the mode i is set. Then, in step §73 and step S74, the variable j and the variable i are initially set to i, respectively, and in step S75, the target pixel data Cdp [i] Lj] is set to 0, and is initially set to zero. Next, in steps S76 and S77, the variable n and the variable m are initially set to 0. In step S 7 8, the following processing is performed: the root pixel data CSP [i + ml] [j + nl] is multiplied by the convolution filter coefficient Cv [m] [n] and added to the target pixel data Cdp [i] [ j]. Next, in step S79, it is determined whether the variable m is smaller than two. In the present case, the variable m = 0, so a yes judgment is made. At step s80, the variable m is only increased by 1, and after setting m = 1, it returns to step $ 78. In step S78, the following processing is performed: only the root pixel data of the right neighbor of the i pixel in the x-axis direction is multiplied by the volume sum of the right neighbor in the x-axis direction, and the over / consideration coefficient C v [m] [η], Add on the same target pixel data 匚 dp [丨] [j] to describe ^^. In step S79, the same processing is repeatedly performed until it is judged that the claw is not smaller than 2 (equal to 2). Then, go to step S81 to determine whether the variable n is smaller than 2 and longer than 2 hours. Then go to step S82. After the variable η has only increased by 1, return to --- C. Please read the Jiang Yishi page on the back) 27- 525078 A7 B7 V. Description of the invention (25) S 7 7, repeated execution of subsequent processing. In step S 81, the above processing is repeatedly performed until it is determined that the variable η is not longer than 2 hours (which is equal to 2). With this, the result of repeated drawing and adding the same target pixel data Cdp [i] [j] by 3 × 3 convolution filter coefficients. That is, by this, the convolution filtering operation of one object pixel is completed. Next ', it proceeds to step S83, and judges whether 丨 is smaller than HMAX-2. If it is small, it proceeds to step S84. After the variable i is only incremented by one, it returns to step S75 and repeatedly executes subsequent processing. That is, the target pixel data is moved one by one in the x-axis direction, and the same processing is performed. In step S83, it is determined that the variable i is not longer than ΗΜΑχ-2 hours (when the determination is equal), and it proceeds to step S85 to determine whether the variable j is smaller than vMaX-2. It is judged that the variable j is smaller than VMAX-2, and the process proceeds to step S86. After the variable "is increased by only i, the process returns to step S74, and the subsequent processes are repeatedly executed. That is, while moving the target pixel data in the y direction, the same process is performed repeatedly in sequence. Then, in step S85, when it is judged > νΜΑχ · 2, the convolution overshoot processing is ended. In this way, the same result H can be obtained as in the case of head i 2 or Figure 9 and Figure 20. H is not a polygon unit, but is traced and processed in pixel units. The generation and processing of the address requires time. The results of the tracing process need to be printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. Therefore, it is preferable to perform tracing processing in polygon units as shown in FIG. 12 or FIG. 19 and FIG. 20. Next, a description will be given of the conical treatment. Here, as shown in FIG. 23, the cone-convolution processing is repeated as follows: the average 値 of the four pixel 邻接 adjacent to each other in the processed image is obtained, and the pixels are arranged at the centers of the 4 pixels. That is, the bilinear interpolation method (bi | jne · inter polation) is used to perform the operation in the vicinity of 4 297 mm. II — 111 I — — — — — — II (Please read the Caution page on the back first) _Scale Applicable to the processing by the Consumer Property Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs of the State of China 525078 A7. The average pixel 发明 of the invention description (26) points can be processed from η X n (n is the power of 2) Get (n / 2) x (n / 2) image data. This process is repeatedly performed, and the 1-pixel data of the last cone apex becomes the pixel data indicating that all the pixels on the bottom surface of the cone are average. When performing this cone filtering process, the main CPU 31 outputs the following tracing command for tracing engine 4 i: int L; / * length of one side of the root zone 氺 / int offset; L =; / * start image The length of one side * offset ^ O; while (L > 1) {Set-Texture—Base (0, offset); / base point setting of the texture area 氺 / offset + = L; Set—Drawing-Rectangle (0, offset); / * Set the base point of the drawing area * / Flat—Texture_Rectangle (0, 0, L / 2, 0, L / 2, B / 2, 0, L / 2, 0.5, 0.5, L + 0.5, 0.5 , L + 0.5, L + 0.5, 0.5, L + 0.5, 1.0); L = L / 2; i This drawing command is represented as a flowchart, as shown in Figure 24, initially 'at step S91 , Set 0 at the beginning of the variable offset. Next, in step S92, a process of setting the base point of the texture area 51 to (0, 0ffset) is performed. That is, as shown in Fig. 25, a base point T (0, 0) is set. Secondly, enter step s93 and increase the variable offset by only L. Then, in step S94, the drawing 29-sheet paper H Chinese Standard (CNS) A4 specification (210 X 297 public love) is set.

(Please read the note on the back TS 1 page first)-Binding · V. Description of the invention (27) The base point T (〇, 〇ffset) of the area 52, Ετ 〇-,) The current situation, as shown in Figure 25 As shown, the base point D (0, L). <, times' In step I S95, the following processing is performed: root (texture

^ (〇.5, 0.5, L + 0.5, 0.5, 5 + 1 + 0.5, 1 + 〇5j〇5 > 1 + J pixels 値 multiplied by 1 'added to the target quadrilateral (〇, 〇 , / 2, 0, ML L / 2'0 'L / 2). That is, from this, the processed image of the bottom conical bottom surface shown in Fig. 23) can be obtained! The upper-level processed image. Next, the process proceeds to step S96, where the variable L is 1/2 of the current value. In step S97, it is determined whether the variable L is greater than i. When the variable L is greater than i, the process returns to step: S92, and the subsequent processing is executed repeatedly. That is, the image data of the third layer is obtained from the second layer. Hereinafter, the same processing is repeatedly performed. In step S97, it is judged that the variable B is not larger than i-dai-ji temple (the variable L is judged to be equal to " the temple), and the tapering process is terminated. Next, the difference processing between frames will be described. In this inter-frame difference processing, as shown in FIG. 26, the difference between the frame image at time t and the frame image at time ㉝ ... is calculated. Thereby, a moving image area can be extracted. That is, in this case, the main CPIMI executes the processing shown in the mud chart for the rendering engine 4 丨. Initially, at step s 丨 0 丨, the drawing engine 4! Corresponds to the command from the main CPU 3 丨, and modality 2 is set as the blending mode. Next, in step S 102, the rendering engine 4i uses the frame image data of the time frame as the target image in the image data input by the video camera 35, and uses the frame image data of the time frame as Root image data. Then, in step S103, the rendering engine 41 executes the following processing ... Subtracting the root quadrilateral pixel 値 from the target quadrilateral pixel 値 and rendering. Frame pixel element of target area

This paper size applies the Chinese national standard (CNSM4 specification (2j0x297 mm) 525078 A7 B7 V. Description of the invention (28) The frame pixel data of the material and the root zone are in the still drawing area, which becomes substantially equal to 4 値As a result, the process of step S103 is performed, and the pixel data size becomes approximately 0. 'In this regard, the pixel data size of the moving area becomes different in the target condition and the root condition. Therefore, in The image data i obtained from the processing result of step S103 becomes a predetermined size other than 0. Therefore, the size of each pixel data of the image data of the difference between the frames can be distinguished from the dynamic area = and the still image. Area. Next, the distance between images will be explained. The distance between images is shown in Figure 28, which shows that the two frames of the image, image A and image B have different degrees. Find the distance between images. In the case of distance, the process of obtaining a difference image between the image A and the image B is performed by a process with a clamp (eUmp). The so-called clamped process means that the smaller than 0 is saturated to 0. , And the process of saturating 値, which is larger than the largest 値, to the maximum 値. Find For the distance between images A and b, the difference between the clamped difference image data of image b from image A and the clamped difference of image A from image B is obtained. The difference image data is added, and the absolute difference image data is obtained by adding them together. For example, assuming that the predetermined pixel data of image A is 13, and the expected pixel data of image b is 13. If it is 20, the AB of AB becomes _7, but the clamped result: its 値 becomes 0. In addition, the BA of 値 becomes 7. As a result, the sum of the two tests becomes 。. In this way, the absolute conversion is obtained. In the case of difference image data, the conical processing of the absolute image data is performed next. As described above, if the conical processing is performed, one pixel of its vertex becomes the processed image (absolute Indifference

Specifications of this paper ⑵〇 X (Please read the notes on the back first>

ϋ nnn 5σ4I ϋ H ϋ n III Printed clothing by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 525078 V. Description of the invention (if the μ range exists, proceed to step S139 and perform the processing. Thereafter, return to step S137, The pixel block of the reaction force 1 is located in step SU8, and it is judged to be obtained in the entire range: the same process. Step S140, by repeatedly performing step SU7 ^ space, when entering the distance between the images, extract the shortest distance Corresponding to the above, most of the above template images are for the target image, but the template is much smaller than the target image, and the pattern matching process is performed when the figure is relatively large. As shown in the flowchart, you can: In the case, in step 5}, select the target image =. Then, in step 8152, trace the template of the edge engine ， to generate a second image. In :: Step, calculate the target image selected in step si51 and in it ;: The distance between the images of the template image of Erjin Λ. Although it is the same as the picture, but: 1: Γ! The same is performed for the distance between the images, and the result is like: "'Γ", as shown in Figure 32 Shown because of the arrangement of most templates Image, so use this template unit to calculate the distance between the images. = Second, if it is determined whether the range of the distance between the images exists in all the search ranges, go to step S155, just execute, go back to step one and repeat: : Step S154 'When it is determined that the distance between the images is obtained in the entire search range, step S156, the shortest distance between the images is extracted from the distance between the images obtained by repeatedly performing the processing of step S 丨 53. Figure 33 shows an example of processing when a motion vector is detected. Initially, it is called 1 in step.

(CNsuTi ^ F Yang Lu Ding-34- (210 x 297 Υ 525078 A7 B7 V. Description of the invention (32 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs' PCRTC 44 makes the previous frame image output by the video camera 35 and The current duck image is stored in the image memory 43 as shown in Fig. 34. Next, it proceeds to step S162 'The main CPU 31 extracts a block of the previous frame image as a template. Then, in step S 1 63, the main CPU 31 Perform the following processing as explained with reference to the flowchart of FIG. 29. The image distance between the template extracted in step S62 and the image (block) corresponding to the current frame is obtained. Next, in step S164, As shown in FIG. 35, the main CPU 31 determines whether to search all within the search range of the current frame. When the range that has not been searched exists, it proceeds to step S165 and executes the following processing: only moves the map corresponding to the range of the pixel-presented frame, Image (block) position. Then, return to step S163 to perform the following processing: find the image distance between the template image and the block image of the current frame. In step S 1 64, repeat the above processing until it is determined that Search within the entire search range. In step S164, it is judged that when searching is performed in the entire search range, the process proceeds to step S166, and the ± cpu 31 obtains the shortest image distance from the distances between multiple images obtained by repeatedly executing step $ ⑹ = processing, and selects this distance. The corresponding block on the current frame is regarded as the shortest block. Then, in step M 6 7, Wang CPU 31 finds the template and the shortest motion vector selected in step S166. ^ Uw Next, enter step u168, ACPU31 #, mM whether the block is Calculate the motion vector. Not yet. When the block of the motion vector and the motion vector left in the 妒 σσ 00 is left, return to y step S 1 62, and extract the same new processing from the previous frame. Change the board, repeatedly execute step S168, and judge that all the processing is completed. 疋 When Sarah turned to I, 35-(210x ^ 97 ^ 57 · (Please read the precautions on the back first Ϊ 1. Installation --- 呙(This page) · 525078 A7 B7 V. Description of the invention (33 As above, the motion vector can be obtained, but for example, as shown in Fig. 36, 'the motion vector can also be obtained. In this example', in step S17l, the video camera is forwarded; The image and the current duck image are stored in the image memory 43. Second, at = L: 2 'such as _38 Γ cat In the step, the main CPU 31 determines whether to move the relative position of the current building to the front quietness within the entire range. The search for moving miscellaneous objects in the entire range When the time is set, the process proceeds to step S174. The main CPU3: = performs the following processing: only shifts the relative position of the gravity i pixel realizing frame with respect to the previous frame, step sm, and executes the following processing: then calculates the distance between images of each block simultaneously; . Brother t1: Repeat the above processing to step sm, and judge that the existing building is moved in Fanyuan. The distance between the images is obtained in step S172 2. Therefore, in the steps of: 刀 刀 木 内 松 μ soil / κ I / * 3, the judgment is within the whole search range = the current building, the number of blocks multiplied by the searched image distance can be obtained. μ 糸 number (sigh "rSm, for one block, choose the shortest, selected current Pabu M 诒 谇 from the distance between image points in the search range, and the short distance between the corresponding short images. : Two Γ shortest blocks. The same processing is performed for each block. S2, S176, each block finds the motion vector of the most difficult sum frame obtained in step 75. Therefore, by obtaining the motion vector, Compared with the figure-find the movement vector more quickly. Ϋ́33 所-compared to the sentiment 'can also be obtained as shown in the flowchart of Figure 38, compared to Figure 33 -36 (210x297 mm)

(Please read the note on the back TE 'install --------- order --------- 呙 this page) #-Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs ^ 5078 V. Description of the invention (The case shown in (34) can also find the motion vector more quickly. That is, in the example of FIG. 38, in step s 丨 8 丨: the previous building and the current paste image are stored in the image recording step S182, The main CPU 31 is as shown in Figure 3 = 43. Secondly, care about the continuation and the image, and create a low-resolution image In step M, the main processor 1 performs the following processing: using the page. Two == management. At: = where to find the motion vector: step = 83 to find the low-precision motion vector, proceed to step si84. : Based on the original resolution (based on cone tapping), each block searches for a low-precision motion vector package, and performs the process of finding this motion vector. After 36 dead weights, Hough transform (Hough transfomat 〇n > To explain the process of transforming a straight line into a point by the human exchange system, the rules are as follows:

Ah (P, θ) = ^ A (x, y) β (p-XCOS0-y sine) d χ dy 04) The above-mentioned wisdom fee is deformed to obtain: θ) ^ / a (p COS0-t sine -p sinG- t cosG) dt (15) FIG. 40 shows a processing example when the rendering engine 41 performs this Huff transform process. Initially, in step S19: [, 0 is set at ㊀. This is printed as shown in Figure 4i (CNSM4 specification: 0 × 297 mmh 525078, printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs, printed A7 B7 5. Invention Description (35)), which indicates the rotation angle of the input image. At step si92, The operation will only input the image data of the angle affine transform (affine transfomation). This angle θ specifies the coefficients 3 to d in the above formula (11). Its / person 'enters the step% S 1 93 'Perform the following processing: 纟 bidirectional reduction in the y-direction (bihnear). The details of the bilinear reduction processing in the y-direction will be described later with reference to the flowchart in FIG. 42. The input image that has been rotated by Θ can be expressed as a y-mu line in the y-axis direction. Next, the process proceeds to step S194, and the following processing is performed: the process in step si% is reduced to a linear point sequence into the drawing area. The angle ㊀ position on W. Next, go to step S195 to determine whether θ is equal to redundant. When not waiting, go to step S196, Θ only increases by π / η, return to step S192, and repeat the subsequent processing. When it is judged that ㊀ and verbose are equal, it ends Processing: With this, for example, when the angular resolution is π / η, a maximum of 6n polygon Hough transforms of 64 x 64 pixels can be used. In addition, in the case of image data of 256 x 256 pixels, the angle When the resolution is Y / n, the maximum polygon Huff transform can be reduced. Next, the bilinear reduction processing in the y direction in step S193 in FIG. 40 will be described in detail. When the rendering engine 41 is executed, The main cpu 3 t supplies, for example, the following command to the engine ^ 4: intL, / * length of one side of the root zone * / int offset · L〇-2λν: / * length of one side of the starting image * / l = = lo ： This book uses Chinese National Standard (CNS) A4 specification⑵GX 297 public love) -------------- install -------- order ----- ---- ^ (Please read the special note on the back? This page first) # · 525078 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the invention (36) offset = 0: while (L >) (Set —Texture_Base (0, offset); / 的 Set the base point of the texture area * / offset + = L; Set_Drawing-Base (0, offset): / * Set the base point of the drawing ^ area * / Flat_Texture— Rectangle ( 0, 0, L0, L / 2, 0, L / 2, 0, 0.5, L0, 0.5, L0, L + 0.5, 0, L + 0.5, 1.0): L ~ L / 2;} \ If this processing table is not “flow chart”, tjt 'is taken as shown in FIG. 42, and in step S20 1, L0 is set at the beginning of the variable L. At the same time, 0 is set at the beginning of the variable offset. Next, in step S202, the following processing is performed: The base point of the texture area 51 is set to (0, offset). That is, as shown in FIG. 43, the base point T (0, 0) is set. Next, in step S203, the variable offset is increased by only L. Then, in step S204, (0, offset) is set at the base point of the tracing area 52. In the present case, as shown in Fig. 38, a base point D (0, L) is set. Next, in step S205, the following processing is performed: the pixel (値) of the root (texture area) quadrilateral (0, 0.5, L0, 0.5, L0, L + 0.5, 0, L + 0.5) is multiplied by 1, and added to The target quadrilateral (0, 0, L0, 0, L0, 1/2, 0, L / 2) is depicted. Next, it proceeds to step S206, where the variable L is 1/2 of the current value. In step S207, it is determined whether the variable L is greater than 1. When the variable L is greater than 1, the process returns to step S202, and the subsequent processing is executed repeatedly. __—_- 39- _ This paper is compliant with ... National Standard (CNS) A4 (210 X 297 mm) — One (Please read the notes on the back first; this page) Binding: 525078 A7 B7 5. Description of the invention (38) times' In step S225, the image of the immediately preceding frame is used as the root image. In step S226 ', the following processing is performed: the source is multiplied by a convolution coefficient, and the target is traced. Next, the process proceeds to step S227, and if it is determined whether all of the n convolution coefficients have been multiplied, the convolution coefficients that have not yet been multiplied exist, the process proceeds to step S228, and the process of changing the convolution coefficients is executed. Then, in step S229, the variable N indicating the number of times the convolution coefficient is changed is increased by only one. In step S230, it is judged whether the number of times of changing the convolution coefficient is larger than η / 2, and the variable N is smaller than η / 2. Then, the process returns to step S225, and the subsequent processes are repeatedly executed. When it is determined in step S230 'that the variable ν is larger than η / 2, the process proceeds to step S231, and the image of the main frame is used as the root frame. Then, the process returns to step S226, and subsequent processes are executed repeatedly. When it is determined in step S227 that all of the n convolution coefficients are multiplied, the process ends. That is, as shown in FIG. 47, when the convolution coefficients are 4 (η = 4), frames Fn and f22 that are close to the immediately preceding frame F1 are generated based on the immediately preceding frame F 1 (the first and second renderings). The generated frames, frames F23, PH (frames generated by the third and fourth depictions) are close to the current frame h, so they are generated based on the current frame. In this case, the direction of the movement vector and the original direction It is in the opposite direction, so the shading process is performed in consideration of this. As shown in FIG. 46, the motion vector exists in units of macroblocks. Therefore, this movement blurring process is performed in units of macroblocks. When the engine 4i executes various processes as described above, for example, the CRT 36 can display the processing results as shown in FIG. 48. In this case, the main cpu3i executes the processes shown in the flowchart in FIG. 4-9. (Please Read the back t notice first,

Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs, the Consumer Cooperative Cooperative -41-525078 Printed by the Consumers' Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs, 42- A7 V. Description of the Invention (39) That is, initially, in step S25 1, the main CPU 3 1 makes use of The video image input by the full-scale 4W, Katana video camera 35 is taken through the PCRTC 44 and the memory interface to make the image memory 43 memorize. Next, in step S252, the main CPU 31 controls the rendering engine. As described with reference to the flowchart of FIG. 27, the inter-frame difference processing (dynamic extraction processing) is executed. 0 Secondly, in step S 2 5 3, the main CPU 3 1 controls the rendering. The engine 41 executes edge extraction processing. 50 is set as the convolution filter coefficient of the convolution filter process described with reference to FIG. 12 and this edge extraction process is performed. Next, the process proceeds to step S254, and the main CPU 31 controls the rendering engine 4; The Huff transform process described in the flowchart of 40 is executed. Furthermore, in step S255, the main CPU 31 controls the rendering engine 41 to execute line segment extraction processing based on the widower conversion processing performed in step M ". In step S256, it is determined whether the user has instructed steps S252 to S255. The processing result shows that when there is no command, it returns to step s25 ^, and executes the subsequent processing. In step S256, it is judged that the command processing result is: when it goes to step S257, the main CPU 31 controls the PCRTC 44 through the memory interface 46 reads out the image drawn in the image memory 43 and outputs it to 36 'for display. Thereby, for example, the image shown in FIG. 48 is CRT 36.' 'In the display example of FIG. 48, in step The image of 8252's _difference processing out processing) is displayed on its far left side, at the edge of step S253: the out processed image is displayed on its right side. Furthermore, on its right side, the input taken in step S25 丨The image is displayed on this paper scale to the Chinese National Standard (Cii) A4 Specification ⑵) I ------------ • Install --- (Please read the note on the back first? ^^^ write (This page) · 525078 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Description (4〇) Take out the image of the Huff transform processed in step S254 (the output image of the eighth ♦ Θ in Figure 41) is displayed on top of it, and the line segment is processed in step Yi Nao. The result (the line segment extracted by the Huff transform) is shown on the upper side. The above description is based on the case where the present invention is applied to a computer entertainment device ', but the present invention can also be applied to other image processing devices. In this manual, the so-called system refers to the entire device made up of a large number of devices. Also, as for the provided media provided to the user by a computer program that performs the above-mentioned processing privately, in addition to the magnetic disk, CD-ROM, In addition to solid-state memory and other green media, communication media such as the Internet and satellites can be used. [Effects of the invention] As described above, according to the image processing device described in item i of the scope of patent application 申请 Patent_item 8 The image processing device method and the patent application No. 9 (provide media, because the following actions are repeatedly performed to obtain the predetermined operation result: the predetermined operation is performed in the pixel unit and the second unit is drawn in the polygon unit. Remember Memory mechanism, so you can perform image processing with a simple structure and a low-cost device. _The image processing device contained in the patent application, the image processing method contained in item 17 of the application scope, and the scope of patent application Since the provided media described in item 8 generates a tracing command, the tracing command is executed repeatedly to obtain a predetermined calculation result: applying the pixel unit to :, and differently using the polygon unit. Precautions can be performed by using the drawing command. According to the image processing device contained in Item 19 of the scope of patent application, the application is 43 丨 China National Standard (CNS) A4 Specification (2l0x $ f ^ jy- ------ ^ ---------. (Please read the note on the back of tf first ^^ write this page} 525078

V. Description of the invention (μ) The image processing method described in item 26 of the scope of the printed product and the provided media in item 27 of the patent application park by the Intellectual Property Bureau of the Ministry of Economic Affairs ’s consumer cooperatives, because Part of the _ in the predetermined operation is drawn in the second memory in polygonal units, and the other part of the predetermined operation is performed in pixel units, and the image data is added or subtracted to be drawn in the second memory in polygons. Mechanism, so by drawing the edge processing can make various calculations quickly execute. [Brief Description of the Drawings] FIG. 1 is a block diagram showing a configuration example of a conventional image processing apparatus. FIG. 2 is a flowchart illustrating a conventional convolution filtering process. FIG. 3 is a diagram illustrating root pixel data. Figure 4 illustrates the convolution filter coefficients. FIG. 5 is a diagram illustrating target pixel data. FIG. 6 is a block diagram showing a configuration example of an image processing apparatus according to the present invention. FIG. 7 is a diagram illustrating a green area inside the image memory 43 of FIG. 1. FIG. 8 is a diagram illustrating bilinear interpolation processing. FIG. 9 is a diagram illustrating a convolution filtering process. Figure 10 is a diagram illustrating the source pixel data. FIG. 11 is a diagram illustrating a convolution filter coefficient. FIG. 12 is a flowchart illustrating a convolution filtering process. Figure 13 illustrates the target pixel data. FIG. 14 is a diagram illustrating target pixel data. Figure 15 is a diagram illustrating target pixel data. Figure 16 is a diagram illustrating target pixel data. Figure Π is a diagram illustrating target pixel data. -------------- Equipment ---- (Please read the V Notice on the back page first) Order --------- #. -44- U5078 Intellectual Property of the Ministry of Economic Affairs Printed by the Bureau ’s Consumer Cooperatives 5. Invention Description (42 Figure 18 is a diagram illustrating the target pixel data. Figure 19 is a flowchart illustrating another example of convolution filtering processing. Figure 20 is a flowchart illustrating another example of convolution filtering processing. Fig. 21 is a diagram illustrating the target pixel data in step S53 of Fig. 19, and Fig. 22 is a flowchart of another example of the convolution filtering process of the alpha element. Fig. 23 is a diagram illustrating the cone filtering process. 24 is a flowchart explaining the cone filtering process. FIG. 25 is a diagram explaining the processes of steps S92 and S94 of FIG. 24. FIG. 26 is a diagram explaining the inter-frame difference process. FIG. 27 is a flowchart explaining the inter-frame difference process. FIG. 28 FIG. 29 is a flowchart illustrating the distance between images. FIG. 29 is a flowchart illustrating the distance between image calculation processing. FIG. 30 is a flowchart illustrating the pattern matching processing. FIG. 31 is a flowchart illustrating another example of the pattern matching processing. FIG. 3 is a flowchart for explaining a pattern matching process. FIG. 3 is a flowchart for explaining a motion vector detection process. FIG. 34 is a diagram for explaining a motion vector extraction process. Fig. 35 is a diagram explaining how to move blocks within a search range to find the reason. Fig. 36 is a flowchart explaining another motion vector detection process. Fig. 37 is a diagram explaining how to obtain the famous name of Fig. 36. The shoulder image of the distance between the images in step ^ S172 is used to explain the other motion vector detection processing. Figure 39 is used to explain the processing of step-s 1 " "Figure in step% S182 of Figure 38." Figure. The place where the moving vector is plotted (please read the note on the back first to write this page). The μ scale is applicable to China-45- 525078

V. Description of the Invention (43) FIG. 40 is a flowchart illustrating Hough transform processing. Figure 41 is a diagram illustrating the Hough transform. Fig. 42 is a flowchart illustrating the y-direction bilinear reduction processing at step s93 in Fig. 40. FIG. 43 is a diagram illustrating the processing of steps S202 and S204 in FIG. 42. FIG. 44 is a diagram explaining a motion blur process. Fig. 45 is a flowchart illustrating a motion blur process. Fig. 46 is a diagram illustrating a ccd moving image correction process with an electronic shutter. Figure 4 7 ^ illustrates the motion blur processing. FIG. 48 is a diagram showing a display image display example. FIG. 49 is a flowchart illustrating image display processing. FIG. 50 is a diagram showing an example of a filter coefficient when performing edge extraction processing. [Description of component numbers] 3 1 main CPU '32 main memory, 33 image processing chip, 34 bus, 35 video camera, 36 CRT, 41 drawing engine, 42 memory interface, 43 image memory, 44 available Programmable cathode ray tube controller, 45, 46 bus, 51 texture area, 52 drawing area.

(Please read the notes on the back Z · I I I I I I 1 · 11111111 ▼ write this page) #. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs -46-

Claims (1)

525078 Patent No. 088 · 144, A8, Chinese Patent Range Amendment (January, 2010) C8 Sixth, the scope of patent application '" ^ ~ ~ ^-1. First, the image processing equipment 2 is characterized by: 2 memory mechanism: the root image data is stored in pixel units;! I S two memory mechanism: The target image data is memorized in pixel units: and, the tracing mechanism: repeatedly performs the following operations to obtain a predetermined operation result: "The opposite image It is described in the first-the memory mechanism of the first memory mechanism in pixel units Those who use predetermined operations to describe the target image data in the polygon unit described above in the unit of a polygon ^. 2. An image processing device according to item i of the patent application scope, wherein the source image memorized by the aforementioned first memory mechanism The data is the image data output by the video camera. 3. The image processing device according to item 丨 of the patent application scope, which further has a designation mechanism: specifies the operation mode between the aforementioned source image data and the aforementioned target image data. 4 · If the image processing device according to item 3 of the patent application scope, wherein as the aforementioned operation modality, the aforementioned designating agency specifies the first modality · · The aforementioned source image data is added to the aforementioned target image data, Or the second mode: subtract the aforementioned source image data from the aforementioned target image data. 5. The image processing device of item 4 in the scope of patent application, wherein as the aforementioned operation mode, the 'specified agency' specifies the first Three modalities: the aforementioned source image data is memorized in the aforementioned second memory mechanism as the aforementioned target image data. 6. The image processing device according to item 1 of the patent application scope, wherein the predetermined operation is a convolution filtering process, a cone Shape filtering, inter-frame difference processing, distance processing between images, Hough transform processing, motion blur processing, or Chinese paper standard (CNS) A4 specification (210 X 297 mm) applicable to this paper standard 525078 A8 B8 C8 D8 Bilinear interpolation processing-processing. 7. If the scope of the patent application is the first broad-spectrum, image processing device, the aforementioned image processing device is a computer entertainment device. An image processing method, The department has a first memory mechanism: to remove w with a picture, to memorize the image data of the root image in the early position; and, a brother and one memory mechanism: to exaggerate the picture with a picture ^ ,,, often The image processing method of the image processing device for memorizing the target image data is characterized in that it includes a drawing step: repeatedly performing the following actions to expect a dare operation result: A pixel unit performs a predetermined operation and a polygon unit is used to describe the person __, and the rain is drawn in the aforementioned second memory mechanism as a target image material. 8. 9. A providing medium, characterized in that: providing an image The processing device executes a program including a drawing step, and the image processing device includes a first storage mechanism that stores image data of the root image in pixel units; and a second storage mechanism that stores the target image in pixel units. Image data, this drawing step repeatedly performs the following operations to obtain a predetermined calculation result: the root image data stored in the first memory mechanism is subjected to a predetermined operation in pixel units and the polygon unit is drawn in the second memory mechanism as Target image profiler. 10 · —An image processing device, characterized by: having a memory mechanism: having a first memory unit: memorizing the roots in the unit of pixels -2- This paper size applies to the Chinese National Standard (CNS) A4 specification (210 X 297 Public love) Binding A8 * B8 Like the data; and, -Fenwei,. ^. Data. A w 己 丨 Sibu · in pixel units? The target image resource generation mechanism has been recalled. # ^ Χ,, is a drawing command, the drawing command is to make the following actions, overlaying to the predetermined operation result: the memory in the aforementioned first- The memory image data is subjected to a predetermined operation in pixel units and a polygon is simply used as the target image data in the second memory section; and, / the mechanism. The person who executes the trace command generated by the foregoing generating mechanism. ^ Shen = the image processing device of the scope of patent No. 10, wherein the root image data stored in the first memory unit is image data output by a video camera. 2. The image processing device according to item 10 of the scope of application for a patent, wherein the aforementioned generating mechanism further specifies an operation mode between the aforementioned source image data and the aforementioned target image data. 13. The image processing device according to item 12 of the scope of patent application, wherein as the aforementioned operation mode, the aforementioned generating mechanism specifies the first mode: adding the aforementioned source image data to the aforementioned target image data, Two modes: Subtract the aforementioned source image data from the target image data. 14. The image processing device according to item 13 of the scope of patent application, wherein, as the aforementioned operation mode, the aforementioned generating mechanism further specifies a third mode: make the root image '贝 料 $ 其 忆 于 纟' "Shidi II * A memory unit serves as the aforementioned target image data. 15. The image processing device according to item 10 of the scope of patent application, wherein the pre-footing operation is a convolution filtering process, a cone filtering process, a difference processing between towels, a nose distance processing between images, and a Huff transform processing. 、 Movement Blur • 3- This paper size applies to China National Standard (CNS) Α4 specification (210 X 297 mm) Binding 525078 ABCD 々, any of patent processing scope processing or bilinear interpolation processing. 16. The image processing device as claimed in claim 10, wherein the image processing device is a computer entertainment device. 17. · An image processing method, which is an image processing method of an image processing device having a memory unit that stores image data, and is characterized in that it includes a memory step: the root image data is stored in pixel units in the first The memory unit simultaneously stores the target image data in pixels and units in the second memory unit; and, a generating step: generating a drawing command, which repeatedly performs the following operations until a predetermined calculation result is obtained ... Bu Yi applies pre-foot calculation to the root image data stored in the first memory unit in pixel units and draws in the second memory unit in polygon units as the target image data. 18. A type 1 supply medium is characterized in that a program for causing an image processing device having a memory section that stores image data to execute a process including a memory step and a generating step is provided. The pixel unit is stored in the first memory unit, and the target image data is stored in the pixel unit in the second memory unit. &Quot; ~ " This generation step generates a drawing command, which makes the following actions repeatedly until the As for the calculation result *: In the aforementioned memory step, the root image data of the first memory portion of the memory $ f is subjected to a predetermined operation in pixel units, and the polygon image is drawn in the second memory portion as the target image data. -4-^ U78 19. An image processing device, characterized by: having ^ _: self-memory mechanism: memorizing root image data in pixel units; object-redundant mechanism: memorizing target image data in pixel units; image 2: memory In the aforementioned first memory mechanism, the "multi-lateral opening" is used in the early stage to perform the -partial operation in the predetermined calculation and the material is K; the early stage is depicted in Lishu's second memory mechanism as the target image negative image data riding mechanism: the memory is stored in the The above-mentioned root mechanism of the memory mechanism mentioned above / the pixel unit applies the other-parts in the predetermined operation, and the image data expected by the polygon = M mechanism is added or subtracted, whichever is greater.-Tian, Yu Do not mention the second memory mechanism as the target image data 20. The image data of the first memory mechanism as described in the first patent application scope. The image processing device of item 19, wherein the source image data stored in is the video camera output 21, • 2: Please request the image processing device in the scope of patent No. 19, which has a mechanism: specify the operation mode between the aforementioned source image data and the aforementioned target image data. Of An image processing device, in which, as the aforementioned operation mode, the aforementioned specifying means designates a first mode: adding the source image data to the target image data, or a second mode: subtracting from the target image data The aforementioned source image data. 23. The image processing device according to item 22 of the scope of patent application, wherein as the aforementioned operation modality, the aforementioned designation agency specifies a third modality ... ® National Standard (CNS) A4 Lin_X Tear mm) Binding 525078 Patent application scope Source image data is stored in the aforementioned image data. Figure 24. For example, the image processing device of item 19 in the scope of patent application, computational silk product processing, "㈣M ^ == arithmetic processing, Hough transform processing, motion blur processing, or double line Any of the interpolation processing. X 25. For example, the image processing device of the 19th patent application scope, the processing device is a computer entertainment device. ',' ° 象 26 · — a kind of image processing method, with a memory Institution: memorize the root image data in pixel units; and, the second memory unit: the image processing device of the image data processing unit of the old standard image data in the image processing unit. & is characterized in that it includes a drawing for the first drawing step: applying the predetermined operation in pixel units to the root ordering data memorized in the aforementioned first-memory mechanism = part drawing = polygonal unit drawing in the aforementioned first The second memory mechanism is used as the target image appreciation, and the first step is described as follows: the root image data stored in the aforementioned first-memory mechanism is subjected to other "partial operations" in the predetermined operation in pixel units, and before The polygonal unit of the rider who has been drawn in the nth step is described in the aforementioned second memory mechanism as the target image data person. 27. A kind of media, characterized in that: providing an image processing device to execute a step including the first drawing step and the second A processing program for rendering steps. The image processing apparatus includes a first memory unit: memorizing root image data in pixel units; and a second memory unit memorizing target image data in pixel units, -6 -This paper size is in accordance with Chinese National Standard (CNS) A4 specification (210X297 mm) 7. Scope of patent application The source image data of the first drawing step is recalled in pixel unit 2 to the aforementioned first memory mechanism and polygonal The unit is interpolated from a part of the calculation data in the second difference, the v hidden mechanism is used as the target image, the second drawing step is the second, the source image data is Tu Xinjun / # 思 于 w 述 第-记忆The mechanism of the Jingjing early package is based on a predetermined calculation, and the other part of the aforementioned Ha Yunyi is described by the polygon early position Yu Lun 4 minus the unexpected. The second memory mechanism is the goal. Target image resources This paper size applies to China National Standard (CNS) A4 (210X297 mm)